biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[V1][Metrics] Add several request timing histograms (#12644)

Browse Source 
Signed-off-by: Mark McLoughlin <markmc@redhat.com>
This commit is contained in:
Mark McLoughlin
2025-02-11 15:14:00 +00:00
committed by GitHub
parent 110f59a33e
commit 75e6e14516
16 changed files with 335 additions and 85 deletions
Download Patch File Download Diff File Expand all files Collapse all files

31
tests/entrypoints/openai/test_metrics.py
View File

@@ -85,6 +85,10 @@ EXPECTED_VALUES = {
"vllm:time_per_output_token_seconds":
[("_count", _NUM_REQUESTS * (_NUM_GENERATION_TOKENS_PER_REQUEST - 1))],
"vllm:e2e_request_latency_seconds": [("_count", _NUM_REQUESTS)],
"vllm:request_queue_time_seconds": [("_count", _NUM_REQUESTS)],
"vllm:request_inference_time_seconds": [("_count", _NUM_REQUESTS)],
"vllm:request_prefill_time_seconds": [("_count", _NUM_REQUESTS)],
"vllm:request_decode_time_seconds": [("_count", _NUM_REQUESTS)],
"vllm:request_prompt_tokens":
[("_sum", _NUM_REQUESTS * _NUM_PROMPT_TOKENS_PER_REQUEST),
("_count", _NUM_REQUESTS)],
@@ -169,6 +173,18 @@ EXPECTED_METRICS = [
"vllm:e2e_request_latency_seconds_sum",
"vllm:e2e_request_latency_seconds_bucket",
"vllm:e2e_request_latency_seconds_count",
"vllm:request_queue_time_seconds_sum",
"vllm:request_queue_time_seconds_bucket",
"vllm:request_queue_time_seconds_count",
"vllm:request_inference_time_seconds_sum",
"vllm:request_inference_time_seconds_bucket",
"vllm:request_inference_time_seconds_count",
"vllm:request_prefill_time_seconds_sum",
"vllm:request_prefill_time_seconds_bucket",
"vllm:request_prefill_time_seconds_count",
"vllm:request_decode_time_seconds_sum",
"vllm:request_decode_time_seconds_bucket",
"vllm:request_decode_time_seconds_count",
"vllm:request_prompt_tokens_sum",
"vllm:request_prompt_tokens_bucket",
"vllm:request_prompt_tokens_count",
@@ -220,6 +236,21 @@ EXPECTED_METRICS_V1 = [
"vllm:time_per_output_token_seconds_sum",
"vllm:time_per_output_token_seconds_bucket",
"vllm:time_per_output_token_seconds_count",
"vllm:e2e_request_latency_seconds_sum",
"vllm:e2e_request_latency_seconds_bucket",
"vllm:e2e_request_latency_seconds_count",
"vllm:request_queue_time_seconds_sum",
"vllm:request_queue_time_seconds_bucket",
"vllm:request_queue_time_seconds_count",
"vllm:request_inference_time_seconds_sum",
"vllm:request_inference_time_seconds_bucket",
"vllm:request_inference_time_seconds_count",
"vllm:request_prefill_time_seconds_sum",
"vllm:request_prefill_time_seconds_bucket",
"vllm:request_prefill_time_seconds_count",
"vllm:request_decode_time_seconds_sum",
"vllm:request_decode_time_seconds_bucket",
"vllm:request_decode_time_seconds_count",
]